FUTURISTIC GROOVES

Collaborative filtering software is changing the way people choose music, books and other things, by helping them find things they like, but did not know about.

Each year, thousands of films are released and tens of thousands of books published. A big city has thousands of restaurants. How does one deal with such abundance? Reading reviews of films, books and restaurants can provide a guide, but there are more reviews than one has the time to read, and you cannot be sure that the reviewer’s taste matches your own. Word-of-mouth recommendations can help in that regard; friends, after all, are often friends because they share similar tastes.

Collaborative filtering starts off by collecting data on individuals’ preferences. This can be an explicit process, by which a user ranks a book (or CD, or restaurant) on a numerical scale, typically on a scale of one to five. It can also be an implicit process - a purchase, for instance, is a clear indication that an individual is interested in the item in question. But implicit measures can also be more subtle; for instance, the amount of time spent viewing a web page, or even just the “clickstream” - the sequence of links clicked on by a person browsing on the web. These different methods can then either be aggregated into a single score, or stored separately to allow more detailed analysis. And sometimes, consumers will be asked to score the same item in different ways - for instance, what one thought of the food at a restaurant, and what one thought of the service.

The benefit of item-item filtering is that this elaborate similarity calculation need only be done infrequently. Then, when a user ranks a new item - by purchasing it, ranking it, visiting its web page, or whatever - the system can simply call up a pre-calculated list of items that are also likely to appeal to that user. This is what allows Amazon to handle over 30m customers and give instant recommendations, even as the list of items that have been ranked by a customer hanges, since merely calling up the web page for a particular book counts as a ranking. All the calculations are done by Amazon’s powerful server, which creates a list of recommended items and seamlessly stitches that list into the next page sent to the user’s web browser, neatly excluding items they have already purchased.

The value of collaborative filtering has, in any case, already been established. It helps people find things they might otherwise miss, and helps online retailers increase sales through cross-selling. Where the user of a search engine is on a solitary quest, the user of a collaborative - filtering system is part of a crowd. Search, and you search alone; ramble from one recommendation to another, and you may feel a curious kinship with the like-minded individuals whose opinions influence your own - and who are, in turn, influenced by your opinions.